- Mar 3, 2017
- 1,680
- 6,185
- 136
There is no mention of IPC in the Tweet, so 1T looks more likelyIs that 1T perf uplift of 20% or IPC uplift of 20%?
These 13% are a far better Application average than 19%. ZEN3s Results were heavily boosted by Gaming. Applications were more like 10-13%. So SPEC already seems to be an outlier with ZEN3.Zen 3 saw a 13% R20 1T bump over Zen 2 Vs a 19% average
So who do we believe? Xino or the persons that also said RDNA3 will trash Ada in both Performance and Efficiency?
There is no mention of IPC in the Tweet, so 1T looks more likely
These 13% are a far better Application average than 19%. ZEN3s Results were heavily boosted by Gaming. Applications were more like 10-13%. So SPEC already seems to be an outlier with ZEN3.
While not in public space, it is my understanding that Intel probably still has better documentation and integration in the pro world. But I have no real knowledge about it.Apple also does this but with their retailers but not to extent that AMD does. After 2-4 months of the September release Best Buy and the rest discount newly released iPhones by $100-$200 dollars.
In Australia it’s the same as well, official third party resellers discount newly released iPhones after the initial launch period. Every few weeks Mac’s are sold at a 10% discount too.
The only difference with Apple is that the MSRP stays the same on Apples site.
Now, Apple does this to prop up sales and to move stock. There is no company in the world that doesn’t do sales. Either internally or externally. In spite of this Apple is considered a premium company because of the brand value that it holds.
I would consider AMD to be a premium company. More-so than Intel and those words I would have not said a decade ago.
What does Intel have in the public space to be considered a premium company?
It’s lost its leadership in every facet of its products in makes. From nodes, to CPUs, to GPUs.
When your competitor is falling behind you, you become the premium seller in that space. Being a premium company is also a matter of perception and in the CPU space AMD gained a good reputation. While prices do play a part in being seen as premium company it’s ultimately the products and their quality/performance that determine a company’s premium status.
Xino has a good track record and that sounds reasonable. Taking into account what we know about the changes introduced in the Zen 5 uarch and Cinebench performance profiling, the performance uplift in Cinebench should be quite a bit lower than the average, so 20% there sounds pretty good.
CineBench R23 1T | + % | CineBench R23 MT | + % | ||
---|---|---|---|---|---|
5950X | 1655 | 25,586 | |||
7950X | 2066.8 | + 25% | 38,291.2 | + 50% | |
Zen 5 16-core | 2480.2 | + 20% | |||
i9-13900KS | 2358.4 | + 5.2% | 40,287.5 |
There is plenty of gold in there already, all the LGA pads ect. Not to mention the Direct Die people complaining.So how about gold electroplating?
It would be a thin enough layer, while still offering the lustre and shine of gold.Wouldnt be too expensive.
Suitable for ultra-high end CPUs like Intel's KS parts.
The above scenario is "real" in the world of a die hard that clings super hard on a dream that ARL will be competitive (somehow). If the dream is dead, they have nothing for 2 more years.In what world SPEC, an industry standard in CPU benchmarking used by every single vendor under the Sun, that encompasses multiple different workloads is an outlier, whereas Cinebench, a benchmark for a very specific render (a workload that sometimes isn’t even done on bloody CPUs) is not?
You can debate whether 9000% in SIR is real or not all day long, but the suggestion above is just BS. Might as well use CPU-Z ST score then, it stresses the CPU very well (no it does not).
I think he meant avg 3?%, so ranging from 30 to 39%.
Fortnite 7%, Passmark 11%, R23 1T 20%, 7-Zip 28%, V-Ray CPU 33%, Metro Exodus 38%, Dolphin Bench 71%, WPrime 86%, Avg 3
The numbers Mason, what do they mean?
That's just IPC.WHERE IS MY 40%+ KEPLER
Yeah the only workload on the list that does.So wprime uses avx512?
Because AMD didn't use any gaming Results in their ZEN2 IPC claims, and ZEN4 IPC claims contained gaming results, but they didn't change anything. ZEN3s Gaming results were mostly this good because of the 8 Core CCX with double unified L3, not necessarily because of the Core itself. If we conpare these results without gaming, we land in the 13-15% Region for ZEN2, ZEN3 and ZEN4.Since when is gaming not a primary DIY workload and why would you discount it other than to skew a conclusion on the basis of cherry picked numbers.
Conveniently ignoring that these AMD's claims were verified on this very resource even without games:Because AMD didn't use any gaming Results in their ZEN2 IPC claims, and ZEN4 IPC claims contained gaming results, but they didn't change anything. ZEN3s Gaming results were mostly this good because of the 8 Core CCX with double unified L3, not necessarily because of the Core itself. If we conpare these results without gaming, we land in the 13-15% Region for ZEN2, ZEN3 and ZEN4.
IPC wise, looking at a histogram of all SPEC workloads, we’re seeing a median of 18.86%, which is very near AMD’s proclaimed 19% figure, and an average of 21.38% - although if we discount libquantum that average does go down to 19.12%. AMD’s marketing numbers are thus pretty much validated as they’ve exactly hit their proclaimed figure with the new Zen3 microarchitecture.
I have put respective CB R23 numbers in the table below, if the leaks about 1T's performance number is 20% better than Zen4, we should be expecting 2480 scores in 1T. The number is about 5% better than Raptor Lake KS. I just put MT numbers for future comparison...
CineBench R23 1T + % CineBench R23 MT + % 5950X 1655 25,586 7950X 2066.8 + 25% 38,291.2 + 50% Zen 5 16-core 2480.2 + 20% i9-13900KS 2358.4 + 5.2% 40,287.5
Not bad numbers. That AVX512 implementation seems to be the real bulldozer.Needs translation:
Fortnite 7%
Passmark 11%
R23 1T 20%
7-Zip 28%
V-Ray CPU 33%
Metro Exodus 38%
Dolphin Bench 71%
WPrime 86%
Avg 3?%.
Yeah the only workload on the list that does.
Nope.I think Dolphin does too.
Most of these are in ZEN4s Benchmarks as well, and even the figures are kinda lining up: Fortnite, Passmark way below average, Cinebench slightly below, 7-Zip exactly average, V-Ray slightly above, Dolphin and WPrime way above average. So expected ZEN5 IPC avg 25-30%.Fortnite 7%
Passmark 11%
R23 1T 20%
7-Zip 28%
V-Ray CPU 33%
Metro Exodus 38%
Dolphin Bench 71%
WPrime 86%
Avg 3?%.
We already saw that AMD makes their claims so that they line up with SPEC and I can't deny that, ZEN3 however doesn't get these results in most other Applications, and since ZEN5s claimed SPEC result seems to be 40%+ it looks like it's the same again.Conveniently ignoring that these AMD's claims were verified on this very resource even without games:
yeah and none of them are javascript which is the selling point.Zen 5 IPC figures are subject to re-testing before announcement.
Needs translation:
Fortnite 7%
Passmark 11%
R23 1T 20%
7-Zip 28%
V-Ray CPU 33%
Metro Exodus 38%
Dolphin Bench 71%
WPrime 86%
Avg 3?%.
Yeah but those aren't really relevant workloads for the most part.The simple average of those results is 36.75%